Quadratically converging rational mean iterations (Q803301)

Several examples of mean iterations (mixing algorithms of means: \(a_{n+1}=M_ 1(a_ n,b_ n),\) \(b_{n+1}=M_ 2(a_ n,b_ n);\) \(n=0,1,2,...;\) in addition to the usual properties of means, also homogeneity \(M_ k(\lambda x,\lambda y)=\lambda M_ k(x,y),\quad k=1,2,\) is supposed) with quadratic convergence \((| a_{n+1}- b_{n+1}| =0(| a_ n-b_ n|^ 2))\) are given, both individual algorithms and classes of them. While some are of the arithmetic-geometric mean, \[ M_ 1(x,y)=(x+y)/2,\quad M_ 2(x,y)=(xy)^{1/2}, \] type, leading to solutions of algebraic differential equations (in that case to elliptic integrals) or even, like that with \[ M_ 1(x,y)=(x+y)/2,\quad M_ 2(x,y)=2xy(x+y)^{-1}, \] have algebraic limits, emphasis is laid on the cases where the limits are hypertranscendental functions, for instance for \[ M_ 1(x,y)=4x^ 2y^ 2(x+y)^{-1}(x_ 2+y^ 2)^{-1},\quad M_ 2(x,y)=2xy(x+y)^{-1}. \] \{The paper has a rather informal tone. The same symbol used for different things, stated (for T) or unstated (for R), is not really confusing. On p. 375, line 13 from below, \(c=0,1\) should stand, not \(c\neq 0,1\).\}